|
Activity Number:
|
257
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #301685 |
|
Title:
|
Use of Spatially Adjusted Bayesian Additive Regression Tree (SBART) Model To Reduce Ecological Fallacy in Health Services Research
|
|
Author(s):
|
Ya-Chen T. Shih*+ and Song Zhang
|
|
Companies:
|
The University of Texas M.D. Anderson Cancer Center and The University of Texas Southwestern Medical Center
|
|
Address:
|
Section of Health Services Research, Dept of Biostatistcs, Houston, TX, 77030,
|
|
Keywords:
|
disparities ; ecological fallacy ; Bayesian additive regression tree model
|
|
Abstract:
|
Area-based statistics (e.g., income at zip code level) are often used as proxy SES variables in research lacking person-level data on SES. This census-based approach may lead to ecological fallacy; we applied the SBART model to explore this issue. The SBART incorporates spatial random effects to BART. It has been shown that compared with the census-based approach, estimates obtained from the SBART were much closer to the true estimates (i.e., from data containing information on person-level SES). We applied the SBART method to examine the racial/ethnic disparities in survival among colorectal cancer patients using the link-SEER-Medicare data, with median household income at zip code level as proxy for SES. The comparison between SBART and census-based methods confirmed our concern of ecological fallacy in that SBART reported a weaker association between race/ethnicity and survival.
|
